Steam turbine vibration fault diagnosis method based on deep neural network and manifold alignment

• Main Authors: BAI YU, ZHOU YANG, JIA RENFENG, MA YUTING, YANG ZHAOHAN
• Format: Patent
• Language: Chinese; English
• Published: 13.08.2021
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; HANDLING RECORD CARRIERS; MEASUREMENT OF MECHANICAL VIBRATIONS OR ULTRASONIC, SONIC ORINFRASONIC WAVES; MEASURING; PHYSICS; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS; TESTING

The invention aims to provide a steam turbine vibration fault diagnosis method based on a deep neural network and manifold alignment, and the method comprises the steps: collecting vibration fault data through a vibration sensor, selecting features and fault types, and further carrying out the standardization processing of original data, thereby facilitating the weighting and training; constructing a deep neural network, extracting abstract features, maintaining an original geometric structure of the data by using a manifold alignment item, predicting categories in a classification layer, obtaining a loss function, finally obtaining an overall objective function, iteratively updating a network parameter training model through a gradient descent method until the maximum number of iterations is reached, and obtaining a final network model, predicting a fault category. According to the method, weighting of different characteristic indexes is facilitated, and the learning process is accelerated. Data complex stru